An Investigation Into Ball Grid Array Inspection Techniques

Oesa STM-261
5.4 Comments On Sample 2
Metallographic sectioning confirmed the bridge at P2/P3, missing balls at R I, K8 and
R8 and non-wetted pads at E I0, L4 and K12. Lower volume joints were also noted at PI
and R2. These may have been caused by the formation of a larger bridge around P2/P3
during reflow, with the P2/P3 bridge dominating, drawing solder away from the joints at
P I and R2. Not all defects on this sample were examined metallographically.
The bridges or shorts were easily identified by all X-ray inspection systems. Similarly,
missing balls were identified by all systems but were more easily identified by the
higher anode voltage systems. With a higher anode voltage, the low density images
created by the reflowed solder paste on the substrate, are lighter showing the missing
balls more easily.
The non-wetted pads were less easily located by perpendicular transmission X-ray
systems. These defects show themselves as lacking the tear-drop shape and also may be
slightly out of normal alignment. Identification of non-wetted pads would be difficult
without the tear-drop pad. Careful analysis of the stored images allows identification of
these defects but some were not identified during real-time analysis. Undoubtedly,
detection of non-wetted pads will be more successful as operators become more
experienced.
The angled transmission X-ray system showed the non-wetted pads with much greater
ease. The poor wetting angle to the pad is clearly shown for example in Figure 37. All
non-wetted pads were found during the real-time inspection.
The laminography system automatically detected the bridges around Ll and P3. The
missing ball at H8 was classified as an insufficient joint, whilst the other three missing
balls were identified as missing. The non-wetted pads at EIO, KI2 and L4 were
classified as misaligned. Additionally, voiding in excess of 10% of joint area was found
at joints at A7, All, 88, 810, C5, E15, F14, G2, H2, K13, L3, L6, L9, LIO, M8, PIO
and P 12. This threshold can be altered by the user. Thirty two apparent false detects
were also registered. Hewlett Packard claim that these could be 'tuned out' if a larger
production volume and more programming time were available. It should also be noted
that as these samples were assembled using rework processes, the joint volume may not
be as uniform as high volume production samples and this could account for some of the
false detects. It should be noted that this investigation was based around one off, unique
test samples, whereas, the 5DX is designed for rapid in line testing of many similar
samples.
34